Three centrally edited nationwide soil maps were published in Hungary between 1953 and 1988. Each of these soil maps has advantages, but serious drawbacks as well. Authors’ hypothesis was that the drawbacks of the individual soil maps are correctable with the help of other soil maps and with ancillary data. Therefore, the oldest soil map was digitized and a study was conducted for the harmonization of data on a 266 km2 area at Keszthely (near Lake Balaton) by using the CHAID classification tree method. CORINE land cover database, digital map of surface geology, digital elevation model and derived slope categories were used as ancillary data.The seven source maps contained 7–38 categories. After the intersection of all seven maps, the resulting file contained more than 50,000 polygons and nearly 14,000 category combinations. A variable — showing the probability of the category combinations in relation to the expected areas — was calculated. This was the target variable for classification by the CHAID method, using categories of the seven original maps as independent variables.0.5% of the total area was grouped into 13 less probable classes, which represent the inaccuracies of the initial maps. 99.5% of the total area was classified into 19 classes and some of them were further subdivided on the basis of the geological map. These classes were interpreted as eight WRB soil categories. The final soil map had much better spatial resolution than any of the initial soil maps, non-soil categories were interpreted as soil categories and spatial accuracy was successfully corrected with the proposed method.